Hi,
As someone whom asked to evaluate those tools in Comverse and I really spent lot of time to do that:
1) There is Purify for Linux and Rational call the hole package includes purifyLT,quantifyLT,... Rational test real time (TRT)
I can send you the comparison table if you want to
Anyway we have some problems with it and I asked for some Rational developer site support(they afraid to come over to Israel right now!!!)
2) Insure++ on Linux is too bugy(probably caused by bad porting!!! ) to work with, so after several weeks we decided to not use it at all.
3) I am going to check the current version of valgrind,the previous version was too "beta".
4) There are more tools such mpatrol,fnccheck,... but they don't seems to be high level products.
I would happy to have your comments
David Avrahami
Project Manager
Development Infrastructures
Email: [EMAIL PROTECTED]
visit the Comverse Linux website< http://linux/ >
> -----Original Message-----
> From: guy keren [mailto:[EMAIL PROTECTED]]
> Sent: Wednesday, July 31, 2002 13:06
> To: Guy Baruch
> Cc: [EMAIL PROTECTED]
> Subject: Re: Valgrind (was: KDE 3.0.2 RPM packages)
>
>
>
> On Wed, 31 Jul 2002, Guy Baruch wrote:
>
> > how does it compare with rational's purify tool ?
> > or any other such commercial tool ?
>
> first - you can try valgrind by yourself. you download it,
> compile it, run
> it (with a flag to increase the number of stack frames to show, and
> another flag to tell it to bypass gcc 2.96 bugs - relevant on
> systems with
> gcc 2.96). then you run a binary application with it.
>
> its not as developed as purify or insure++, but:
>
> 1. for linux, there is no purify. there is insure++.
> 2. purify is available for other platforms - valgrind only
> for linux on
> x86 (for now).
> 3. the multi-threaded support in valgrind is not necessarily
> complete -
> unless something changed in the past 3-4 weeks (and it does seem to
> develope very fast lately).
> 4. valgrind has some 'profiling' support in it. purify does
> nto - you need
> to purchase quantify (another rational product for another few K$).
> ofcourse, quantify is more complete then valgrind in the profiling
> aspect (at least it seems so - i didn't try the profiling part of
> valgrind - only read what it can give).
> 5. with valgrind, you don't need to compile nor link the application
> differently then you do for a regular run - you will
> ofcoruse want to
> compile it with debug info, which you do anyway (you do -
> don't you?).
> with purify - you need to link the application with the
> purify tool -
> but you compile it regularly.
> with insure++ - you need to both compile and link the code with
> insure++. they claim there is a mode in which you only link the
> application with insure++, and then you get less info then if you
> compile the code with it. i didn't try this 'linking only'
> mode - the
> documentation was iased towards a full compile with insure++.
> 6. insure++ had some problems with various libraries - i
> needed to tweak
> the linking part in order to get the application i tried
> it with (few
> 10Ks of lines).
>
> in general, you should note that you don't use any of these
> tools as often
> as you compile the application, so it doesn't matter if you
> need to spend
> a _little_ more time on setting things up. this is why i
> think that the
> need to link with purify is not so bad - its the compilation part that
> takes most of the time. linking, even for large projects, usually
> completes within several minutes, and since purify caches the
> libraries it
> instruments, this is usually not that bad. on the first time of
> instrumentation - especially for c++ code - this could be alot. yet,
> valgrind does not cache anything, and needs to re-instrument
> on every run.
>
> --
> guy
>
> "For world domination - press 1,
> or dial 0, and please hold, for the creator." -- nob o. dy
>
>
> =================================================================
> To unsubscribe, send mail to [EMAIL PROTECTED] with
> the word "unsubscribe" in the message body, e.g., run the command
> echo unsubscribe | mail [EMAIL PROTECTED]
>
